From: Angus Dorbie (dorbie++at++sgi.com)
Date: 03/21/2000 10:35:36
I suspect that billboarding wouldn't be a very optimal approach to this.
Ideally you don't want a transformation node per sphere and doing the
vertex transformation for billboards on many complex spheres would be a
huge overhead. The hardware sprite mechanism may come to the rescue on
iR.
What I'd suggest is doing a host/application based billboarded vertex
transformation of a few geometries, if you have orthographic projection
then just one will do, but for perspective you'll need several to cover
the different view vector angles. Maybe you'd render slightly more than
a hemisphere to reduce this work by reducing the number of unique
billboards you need to give an accurate silhouette. Maybe you could
cache lot's of different pretransformed versions of the sphere. Anyway,
as you move through the scene or move the mollecule you'll have to
switch the geometry associated with each atom or particle.
I'd also suggest that you can tesselate more for lighting where you
predict the phong highlight will be on the sphere (or use texture) and
tesselate for a detailed silhouette of the sphere. I don't think
intersections require as much tesselation as the shilouette and you may
not even have any sphere-sphere intersections, in which case a pure
sprite might work well if you can light with a texture.
Cheers,Angus.
Allan Schaffer wrote:
>
> On Mar 19, 7:59pm, Simon Hayhurst wrote:
> > From some work we did a few years ago in SGI Europe, we also found a huge
> > gain by only using hemi-spheres, and rotating the hemi-sphere to face the
> > user. There's a fair bit of App side workload in doing this, but since
> > molecules are typically moving anyway, the net win was considerable.
>
> Yes this would be a tremendous gain. Place the hemisphere geometry
> under pfBillboard nodes so that they'll automatically rotate to face
> the viewer.
>
> Someone might be thinking, "what's the difference, why not just let
> automatic backface culling remove the back-facing side of the
> spheres?": It's because the backface calculations are done on a
> per-triangle basis down in the graphics pipe. Not sending the
> geometry down in the first place would be a big performance win.
> And on MP systems, the billboard calculation is done in the CULL,
> where there's quite typically plenty of extra time each frame.
>
> Allan
>
> --
> Allan Schaffer allan++at++sgi.com
> Silicon Graphics http://reality.sgi.com/allan
> -----------------------------------------------------------------------
> List Archives, FAQ, FTP: http://www.sgi.com/software/performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
-- For Performer+OpenGL tutorials http://www.dorbie.com/"In the middle of difficulty lies opportunity." --Albert Einstein
This archive was generated by hypermail 2b29 : Tue Mar 21 2000 - 10:35:52 PST